Lead/Presenter: Franya Hutchins, CHERP
All Authors: Hutchins F (VA Center for Health Equity Research and Promotion, VA Pittsburgh Health Care System), Rosland AM (VA Center for Health Equity Research and Promotion, VA Pittsburgh Health Care System) Zhao X (VA Center for Health Equity Research and Promotion, VA Pittsburgh Health Care System) Zhang H (VA Center for Health Equity Research and Promotion, VA Pittsburgh Health Care System) Thorpe J (VA Center for Health Equity Research and Promotion, VA Pittsburgh Health Care System)
Objectives:
Latent class analysis (LCA) is a popular empiric approach for describing clinically-distinct groups among medically complex patients, and is used to design programs for high-risk patient populations. However, patients seeking care in multiple healthcare systems may have missing diagnoses across systems, leading to misclassification. We evaluated the impact of multi-system use on the accuracy and composition of patient multimorbidity groups among Medicare-eligible, high-risk Veterans in the Veterans Health Administration (VA).
Methods:
Eligible patients were VA primary care users ages 65 and older, and in the top decile of predicted one-year VA hospitalization risk in 2018 (n = 558,864). Diagnoses of 26 chronic conditions coded in VA encounters and Medicare claims over the previous 24 months were input into latent class analysis models. In a random 10% sample (n = 56,008), we compared the resulting model fit, class profiles, and patient assignments from models using VA-only data versus VA plus Medicare data.
Results:
Over half (57%) of patients had some Medicare-billed care in addition to their VA care. For each chronic condition observed, fewer than 50% of the diagnoses reported in at least one health system were recorded in both. Using a six-class model, we labeled groups based on prevalent diagnoses: Substance Use Disorders (7% of patients), Mental Health (15%), Heart (22%), Diabetes (16%), Malignant Tumor (14%), and High Complexity (10%). The remaining 16% of patients were classified as “unassigned†due to low match probability to any group. The addition of Medicare data improved model fit statistics and decreased the number of “unassigned†patients to 12%. Over 70% of patients assigned to the Substance, Mental Health, High Complexity, and Malignant Tumor groups using VA systems data were assigned to the same group when Medicare data were added. However, 42% of the Heart group and 15% of the Diabetes group were assigned instead to a new group characterized by multiple cardiometabolic conditions.
Implications:
Older adult high-risk patients in the VA healthcare system were sorted into clinically-useful groups based on chronic condition diagnoses using empirical clustering. The addition of diagnoses reported in Medicare improved model accuracy and altered the clinical profiles of groups.
Impacts:
Accessing or accounting for multi-system data will be key to the success of group-tailored interventions for high-risk, medically-complex Veterans.